MobileNet-YOLO Caffe
A caffe implementation of MobileNet-YOLO detection network , train on 07+12 , test on VOC2007
Network
mAP
Resolution
Download
NetScope
Inference time (GTX 1080)
Inference time (i5-7500)
MobileNetV2-YOLOv3
70.7
352
217 ms
inference time was log from script , does not include pre-processing
the benchmark of cpu performance on Tencent/ncnn framework
the deploy model was made by merge_bn.py, set eps = your prototxt batchnorm eps
old models please see here
This project also support ssd framework , and here lists the difference from ssd caffe
Multi-scale training , you can select input resoluton when inference
Modified from last update caffe (2018)
Support multi-task model
Update
CODE UPDATED FOR OPENCV 3
Channel pruning
CNN Analyzer
Use this tool to compare macc and param , train on 07+12 , test on VOC2007
network
mAP
resolution
macc
param
pruned
MobileNetV2-YOLOv3
0.707
352
1.22G
4.05M
N
MobileNetV2-YOLOv3
0.702
352
1.01G
2.88M
Y
0.709
304
1.2G
5.42M
N
0.68
300
1.21G
5.43M
N
MobileNetV2-YOLOv3 and MobilenetV2-SSD-lite were not offcial model
Coverted TensorRT models
YOLO Segmentation
Windows Version
Oringinal darknet-yolov3
test on coco_minival_lmdb (IOU 0.5)
Network
mAP
Resolution
Download
NetScope
yolov3-spp
59.8
608
Model VisulizationTool
Supported on Netron , browser version
Build , Run and Training
License and Citation
Please cite MobileNet-YOLO in your publications if it helps your research:
@article{MobileNet-YOLO,
Author = {eric612 , Avisonic , ELAN},
Year = {2018}
}
Reference
Cudnn convolution